Analysis of thermal performance and ultrasonic wave power variation on heat transfer of heat exchanger in the presence of nanofluid using the artificial neural network: experimental study and model fitting

The present study has experimentally and numerically investigated the influence of ultrasonic waves and temperature variations on the heat transfer coefficient (HTC) by using CNT-water nanofluid. According to the test results, applying the ultrasonic waves in two powers of 35-50 W improves the heat transfer coefficient. The heat transfer coefficient is also enhanced with increasing the concentration of the nanofluid and fluid temperature. The Nusselt number also improved by 540.23% with increasing fluid temperature from 25 to 35 & DEG;C, the mass fraction from 0.12 to 0.25, the flow rate, and the application of ultrasonic waves. Furthermore, the Artificial Neural Network, ANN, is used to evaluate the heat transfer coefficient and Nusselt number. The employed ANNs have the feed-forward architecture with hidden tansig and output purelin neurons. The training method for this network is the backpropagation. Different number of hidden layer neurons as well as dissimilar training method are investigated to find the best ANN performance. Examining the results, it can be seen that the best combination for the present problem is an ANN with 15 hidden neurons and a trainbr training algorithm. Comparing the experimental and ANN results reveals an excellent consistency.

Automated summary

In this study, the influence of ultrasonic waves and temperature variations on the heat transfer coefficient (HTC) using CNT-water nanofluid was investigated experimentally and numerically. The results showed that applying ultrasonic waves at powers of 35-50 W improved the heat transfer coefficient. Increasing the nanofluid concentration and fluid temperature also enhanced the HTC. The Nusselt number increased by 540.23% with increasing fluid temperature, mass fraction, flow rate, and application of ultrasonic waves. Furthermore, an Artificial Neural Network (ANN) was used to evaluate the HTC and Nusselt number, and the results were consistent with the experiments.